Comparing different algorithms that generate phosphene images for visual cortical prosthesis

Keywords
Loading...
Thumbnail Image
Issue Date
2019-06-01
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Visual cortical prosthesis (VCPs) are currently in development and there has already been some speculation about which algorithm is best to use in these implants. Semantic segmentation seems like a obvious choice because the algorithm also gives vision to self-driving cars. However, semantic segmentation is a slow and complex algorithm that predicts around 200 classes, which are not necessary when predicting phosphenes. Therefore, semantic segmentation is compared to two simpler algorithms, edge detection and a proposed neural network that predicts phosphene images from input images. The results indicated that although being the slowest of the methods, semantic segmentation makes the best phosphene images. The edge detection and neural network algorithms need some alterations to be able to make phosphene images that are usable in VCPs.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen